Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Aznatkoiny/zAI-Skills --skill prompt-optimizergit clone --depth 1 https://github.com/Aznatkoiny/zAI-SkillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/aznatkoiny/zai-skills/prompt-optimizer)<a href="https://agentmods.dev/skills/aznatkoiny/zai-skills/prompt-optimizer"><img src="https://agentmods.dev/badge/skills/aznatkoiny/zai-skills/prompt-optimizer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/aznatkoiny/zai-skills/prompt-optimizer"><img src="https://agentmods.dev/badge/skills/aznatkoiny/zai-skills/prompt-optimizer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00205 | $0.02005 |
| Opus 5.5 | $0.00082 | $0.00802 |
| Sonnet 5.5 | $0.00041 | $0.00401 |
| Haiku 4.5 | $0.00020 | $0.00200 |
Grade A, and why
prompt-optimizer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Prompt Optimizer
Optimize prompts for Claude's latest models — the Claude 5 family (Opus 5, Sonnet 5, Fable 5) plus the still-active Opus 4.8, Sonnet 4.6, and Haiku 4.5 — using Anthropic's official guidance. Diagnose what is wrong, pick the right lever, apply a proven fix, then measure.
Two levers, not one
Many problems people try to fix by editing prompt wording are better fixed with an API parameter. Always weigh both:
- Prompt — wording, structure, examples, role, explicit scope.
- Parameters —
effort(intelligence vs. speed/cost),thinking(adaptive on/off),max_tokens(room to think and act).
On the latest models, effort is frequently the highest-leverage change. If reasoning is shallow on a hard task, raise effort to high/xhigh instead of prompting around it. If the model overthinks or over-explores, lower it. Detail in references/model-tuning.md.
What changed for the latest models
These shifts change how to optimize. Do not carry over old habits blindly:
- Effort governs depth, tool use, and thoroughness. Start
xhighfor coding/agentic work, minimumhighfor intelligence-sensitive work. Lower effort means more literal, more scoped, less tool use. - Thinking defaults changed with Claude 5. On the Claude 5 family thinking is on by default (Fable 5 rejects an explicit
disabled); on Opus 4.8 it stays off unless you setthinking: {type: "adaptive"}. Opus 4.6 and Sonnet 4.6 use adaptive thinking.budget_tokensis removed on Claude 5 / Opus 4.8 — control depth througheffort. - Literal instruction following. The model does what you said, not what you implied. State scope explicitly ("apply to every section, not just the first one").
- Dial back anti-laziness prompting. Aggressive "CRITICAL / You MUST" language needed for older models now causes overtriggering. Use normal phrasing ("Use this tool when...").
- Prefilled assistant responses are no longer supported on 4.6+ models (the request returns a 400). Migrations in references/formatting.md.
- Verbosity is calibrated to task complexity and prose is more direct. Tune with positive examples of the concision you want, not "don't" rules.
- Subagent appetite differs by model. Opus 4.8 under-delegates; Claude Opus 5 delegates readily. Steer explicitly — and cap fan-out on Opus 5.
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 90 lines · 205 tokens per session scan A aa1743e22f51
prompt-optimizer is a skill published in the GitHub repository Aznatkoiny/zAI-Skills (10 stars, last pushed 2mo ago), licensed MIT. It adds 205 tokens to every session and 2,005 once invoked, about $0.0008 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
llm-app-development
LLM app development with RAG, prompt engineering, vector databases, and AI agents.
cortex-prompt
Build a production-ready prompt package — system prompt, few-shot examples, output format, edge case handling, eval criteria. Use when asked to "prompt engineering", "build a prompt", "write a system prompt", or "improve this prompt".
meta-prompting
Enhanced reasoning patterns via slash commands (/think, /verify, /adversarial, /edge, /compare, /confidence, /budget, /constrain, /json, /flip, /assumptions, /tensions, /analyze, /trade) or natural language ("argue against", "what could break", "show reasoning", "deep review", "meta-prompts", "thinking modes"…
partner-anthropic
Anthropic intelligence — Claude model family, Claude Code, MCP protocol, Claude for Work programs, May 2026 state, Frank's relationship state, integration patterns. Use when generating Anthropic content, writing about Claude/MCP/Claude Code, partnership-conversation prep for Anthropic, or evaluating new Anthropic…
prompt-hub
Compose the 13-agent Prompt Hub team via @prompt-conductor. Use when designing prompts, optimizing prompts, evaluating prompts, importing patterns from Fabric / awesome-chatgpt-prompts / awesome-claude-prompts, running IFS introspection or psychometric profiling, or building knowledge-base prompt sets. Auto-trigger on…
cost-aware-llm-pipeline
Pick the smallest model that can finish the task, and record the class of work. Use when routing a search, a multi-file edit, or an architecture review, or when a weekly model cap is close.